Clinic Projects

Please click on a link below to view the Harvey Mudd College Computer Science Clinic projects for the corresponding time period.

Clinic Projects for 2013-2014

Augmented Reality in Mobile Devices Powered by Composed Data Services

Client
Aerospace Corporation

Faculty Advisor
Professor Mike Erlinger

Student Team
Sean Adler, Alejandro Frias, and Vivian Wehner

The Aerospace Corporation developed the Enterprise Data Objects (EDO) system to efficiently compose information from multiple data services. Pairing a mobile augmented reality application with EDO backed web services proves useful in understanding situations reliant on geospatial data. We created an application backed by this data, which provides an intuitive way for users to understand their surroundings.

Automatic Fission Track Detection and Analysis in Apatite Crystals

Client
Apatite to Zircon, Inc.

Faculty Advisor
Professor Zachary Dodds

Student Team
Lena Reed, Ravi Kumar, Buike Ndefo-Dahl, and Kevin Vigue

Our project seeks to design and implement algorithms for identifying features within microscopic images of apatite crystals. We built on the work of two previous clinics and refined their techniques to better identify etch pits and fission tracks - two features of interest found with apatite crystals. To this end, we utilized C++ and OpenCV to create modular tools employing pattern matching and custom algorithms. Additionally, we created a LabVIEW interface that fits our tools to our sponsor's workflow.

Collaborative Multimedia Sketchpad

Client
Blackberry

Faculty Advisor
Professor Ben Wiedermann

Student Team
Prachie Banthia, Seo Park, Younji Jeong, and Alistair Dobke

Our liaisons at BlackBerry are interested in the BlackBerry experience for developers building collaborative applications for users in close proximity to one anther. To this end, we are building collaborative applications ourselves. Further, we are building a cloud-based data structure to enable developers to more easily create these kinds of applications. We plan to open-source and publish our applications on BlackBerry World and release the data structure as a package to other BlackBerry developers.

Travel Search Through Images

Client
Expedia, Inc.

Faculty Advisor
Professor Daniel Zimmerman

Student Team
Alana Shine, Jake Low, Haley Erickson, and Jessica Schroeder

The Expedia team has created a vacation search Web application that allows customers to search for destinations implicitly by selecting evocative images. The application presents images in an iterative selection process, culminating in a ranked list of vacation recommendations. Vacation recommendations are derived from the attributes to actual vacation destinations using data derived from traveler review text analysis.

Development and Deployment of Online Physics and Computer Science Courses for Pre- College Students

Client
HMC Online

Faculty Advisor
Professor Mike Erlinger

Student Team
Thomas Ashmore, Sorathan (Tum) Chaturapruek, Corinne Druhan, and Bridgette Eichelberger

Soaring higher education tuition catalyzed the development of MOOCs (Massively Open Online Courses). The HMC Online clinic team created an online platform, MuddX, to host two such courses. MyCS teaches Computer Science to middle school teachers, so they can introduce CS concepts into their classrooms. The other, How Stuff Moves, is based on Mudd's Physics 24. Our team will discuss our platform, the courses, and the plethora of challenges we (and all MOOCs) encounter.

Implementation of Selected Vision Algorithms Using Parallel JavaScript on HSW Graphics

Client
Intel Corporation-Global (collaboration with National University of Singapore (NUS))

Faculty Advisor
Robert M. Keller (HMC) and Wee-Kheng Leow (NUS)

Student Team
John Brooks (HMC), Zheng Duan (HMC), Tiffany Lim (HMC), Nguyen Hien Linh (NUS), Nguyen Truong Duy (NUS), and Nhu Dinh Tuan (NUS)

Intel Labs' Parallel Java Script is a data-parallel programming model that allows JavaScript programs to use hardware parallelism such as multiple cores, vector instructions, and GPUs, which have previously been inaccessible to Web applications. In order to demonstrate the advantages of Parallel JavaScript, we have used it to create a fast, reusable, open-source computer vision library. We then used this library to build a real-time, interactive hand gesture detection application.

Data Visualization for Software Developers of the 21st Century

Client
Intel Corporation-Phoenix (joint with Math)

Faculty Advisor
Professor Daniel Zimmerman

Student Team
Jason Wang, Miranda Parker, Rohitashwa Bagaria, and Sneha Viswanathan (F)

Intel Corporation is interested in creating tools that allow software developers to easily assess the performance or correctness of threaded programs. First, the clinic team explored ways to visualize aspects of threaded execution that would allow developers to easily diagnose performance or correctness issues. Subsequently, the team devised intuitive visualizations of these programs using data collected during program run. The project's end product is an Eclipse SDK plug-in that encapsulates proof-of-concept versions of the team's visualization ideas.

Concurrent Garbage Collection Using Haswell Transactional Memory

Client
Intel Corporation-Santa Clara

Faculty Advisor
Professor Melissa O'Neill

Student Team
Joe Agajanian, Claire Murphy, Will Newbury, and John Sarracino

Intel recently released the Haswell processor architecture. Haswell provides many new features, one of which is hardware support for a parallel execution model called transactional memory. In principle, hardware transactional memory can significantly improve the performance of parallel code. Our goal was to show performance gains in practice by improving the performance of garbage collection, a core technology at the heart of many computer systems.

Algorithms for Mapping Optimization Problems onto an Adiabatic Quantum Computer

Client
Lockheed Martin (joint with Physics)

Faculty Advisor
Professor Theresa Lynn and Professor Jim Boerkoel

Student Team
Taylor Brent, Tessa Adair, Joel Ornstein, and Sean Campbell

Quantum computation has the potential to revolutionize computing by taking advantage of quantum effects like superposition and tunneling. However, designing algorithms that can run on the unusual architecture of these machines is a major challenge. The goal of this project is to investigate methods for mapping optimization problems into a form that can be solved using Lockheed Martin's D-Wave quantum annealing machine. The team is evaluating the performance of heuristic graph mapping algorithms and studying methods for improving solution quality.

Content Based Image Retrieval (CBIR)

Client
MITRE Corporation

Faculty Advisor
Professor Zachary Dodds

Student Team
Aric Hunter, Marissa Novak, Byron Callan, and Neal Kemp

With the increase in tattoos over the past few decades, the biometrics team at MITRE is interested in developing reference software for tattoo recognition and methods for objective performance evaluation. The 2013-2014 clinic team has developed a flexible, simple, and efficient framework for tattoo-matching that uses a variety of computer vision algorithms. This framework enables quick exploration by comparing the relative accuracy of individual algorithms and their combination, and tuning and optimizing the parameters of individual algorithms against representative datasets.

Highly Protective Email Client

Client
Proofpoint, Inc

Faculty Advisor
Professor Geoff Kuenning

Student Team
Mark Mann, Abigail Gregory, Margaret O’Keefe, and Jason Yu

Proofpoint provides email security services to a multitude of corporations. Specifically, Proofpoint filters out dangerous emails before delivery. Some emails could be suspicious, but Proofpoint must deliver them in case they are false alarms. Unfortunately, there is currently no way for recipients to determine which emails Proofpoint suspects. These emails could be dangerous and harm recipients' computers when read. The Proofpoint clinic team built an Outlook plugin that distinguishes potentially threatening emails to inform users' reading decisions.

Application-Layer DASH Client

Client
QUALCOMM

Faculty Advisor
Professor Colleen Lewis

Student Team
Cecily Hunt, Justin Lim, Vijay Ramakrishnan, and Linnea Shin (F)

Qualcomm Incorporated is a leading proponent of the MPEG-DASH standard for the streaming of media content over HTTP. Adoption of the standard has been slow, in part due to a lack of existing media players that support DASH. We created an application-layer DASH client and media player in the form of an Android app in order to increase adoption of MPEG-SAH and support future Qualcomm research.

Respectful Cross-Device Conversion Attribution

Client
Rubicon Project

Faculty Advisor
Professor Colleen Lewis

Student Team
James Kaplan, Adam Blezberg, Mars Park, and Chet Weger

Rubicon Project operates a fully automated trading platform for the buying and selling online advertising, reaching 97% of the U.S. Internet audience. We are designing the building a system that will allow Rubicon Project to link devices that belong to the same person while respecting privacy. Our solution integrates with Rubicon Project's internal infrastructure and will need to be robust enough to handle the massive inflow of 7 billion new data points created daily.

VMware Performance Visualization

Client
VMware, Inc.

Faculty Advisor
Professor Melissa O'Neill

Student Team
Jane Hoffswell, Minnie Lai, Frank Liu, and Rebecca Thomas

This project aims to create a data-visualization tool that developers at VMware can use to understand the performance of complex software systems. Our web-based tool provides a range of customizable visualizations, each developed with unique interactions to allow for flexible exploration of the data. Our system allows users to interact with multiple visualizations and statistics at a time, allowing for maximum comparative ability when extracting correlations from the data.